One-sided MEWMA control charts

In quality control the out-of-control detection problem is often one-sided in the sense that quality decreases as one or more quality characteristics increase. Similarly in environmental monitoring the quality assessment is essentially one-sided. Multivariate control charts like Hotelling's T2 or Lowry's quadratic MEW MA are based on direction invariance. For the one-sided multivariate problem I propose a new detector based on the restricted MLE applied to the MEWMA transform. Moreover I used the Monte Carlo method in order to estimate the operational critical values to implement the control chart for the bivariate case and to give evidence of the full valuable gain in the mean detection delays (ARL) of the new detector with respect to the quadratic MEWMA detector and to the linear CUSUM.